我有一个Python命令行程序,需要一段时间才能完成。我想知道完成跑步所需的确切时间。

我看过timeit模块,但它似乎只适用于小代码片段。我想给整个节目计时。


在Linux或Unix中:

$ time python yourprogram.py

在Windows中,请参阅StackOverflow问题:如何在Windows命令行上测量命令的执行时间?

对于更详细的输出,

$ time -v python yourprogram.py
    Command being timed: "python3 yourprogram.py"
    User time (seconds): 0.08
    System time (seconds): 0.02
    Percent of CPU this job got: 98%
    Elapsed (wall clock) time (h:mm:ss or m:ss): 0:00.10
    Average shared text size (kbytes): 0
    Average unshared data size (kbytes): 0
    Average stack size (kbytes): 0
    Average total size (kbytes): 0
    Maximum resident set size (kbytes): 9480
    Average resident set size (kbytes): 0
    Major (requiring I/O) page faults: 0
    Minor (reclaiming a frame) page faults: 1114
    Voluntary context switches: 0
    Involuntary context switches: 22
    Swaps: 0
    File system inputs: 0
    File system outputs: 0
    Socket messages sent: 0
    Socket messages received: 0
    Signals delivered: 0
    Page size (bytes): 4096
    Exit status: 0

Python中最简单的方法:

import time
start_time = time.time()
main()
print("--- %s seconds ---" % (time.time() - start_time))

这假设程序运行至少需要十分之一秒。

打印:

--- 0.764891862869 seconds ---

import time

start_time = time.clock()
main()
print(time.clock() - start_time, "seconds")

time.clock()返回处理器时间,它允许我们仅计算此进程使用的时间(无论如何,在Unix上)。文档中说“无论如何,这是用于Python基准测试或计时算法的函数”


我将这个timing.py模块放入我自己的站点包目录中,并在模块顶部插入导入计时:

import atexit
from time import clock

def secondsToStr(t):
    return "%d:%02d:%02d.%03d" % \
        reduce(lambda ll,b : divmod(ll[0],b) + ll[1:],
            [(t*1000,),1000,60,60])

line = "="*40
def log(s, elapsed=None):
    print line
    print secondsToStr(clock()), '-', s
    if elapsed:
        print "Elapsed time:", elapsed
    print line
    print

def endlog():
    end = clock()
    elapsed = end-start
    log("End Program", secondsToStr(elapsed))

def now():
    return secondsToStr(clock())

start = clock()
atexit.register(endlog)
log("Start Program")

如果程序中有重要的阶段,我也可以在程序中调用timing.log。但仅包括导入计时就可以打印开始和结束时间,以及总运行时间。(请原谅我晦涩难懂的secondsToStr函数,它只是将浮点秒数设置为hh:mm:ss.sss格式。)

注意:上述代码的Python3版本可以在这里找到。


我很喜欢保罗·麦奎尔的答案,但我使用的是Python 3。因此,对于感兴趣的人来说:这里是他在*nix上使用Python 3的答案的修改(我想,在Windows下,应该使用clock()而不是time()):

#python3
import atexit
from time import time, strftime, localtime
from datetime import timedelta

def secondsToStr(elapsed=None):
    if elapsed is None:
        return strftime("%Y-%m-%d %H:%M:%S", localtime())
    else:
        return str(timedelta(seconds=elapsed))

def log(s, elapsed=None):
    line = "="*40
    print(line)
    print(secondsToStr(), '-', s)
    if elapsed:
        print("Elapsed time:", elapsed)
    print(line)
    print()

def endlog():
    end = time()
    elapsed = end-start
    log("End Program", secondsToStr(elapsed))

start = time()
atexit.register(endlog)
log("Start Program")

如果你觉得这很有用,你仍然应该投票给他的答案,而不是这一个,因为他做了大部分工作;)。


您可以使用Python分析器cProfile来测量CPU时间,以及每个函数内部花费的时间以及每个函数被调用的次数。如果您想在不知道从哪里开始的情况下提高脚本的性能,这非常有用。对另一个堆栈溢出问题的回答很好。查看文档总是很好的。

以下是如何从命令行使用cProfile评测脚本的示例:

$ python -m cProfile euler048.py

1007 function calls in 0.061 CPU seconds

Ordered by: standard name
ncalls  tottime  percall  cumtime  percall filename:lineno(function)
    1    0.000    0.000    0.061    0.061 <string>:1(<module>)
 1000    0.051    0.000    0.051    0.000 euler048.py:2(<lambda>)
    1    0.005    0.005    0.061    0.061 euler048.py:2(<module>)
    1    0.000    0.000    0.061    0.061 {execfile}
    1    0.002    0.002    0.053    0.053 {map}
    1    0.000    0.000    0.000    0.000 {method 'disable' of '_lsprof.Profiler objects}
    1    0.000    0.000    0.000    0.000 {range}
    1    0.003    0.003    0.003    0.003 {sum}

我喜欢datetime模块提供的输出,其中时间增量对象以人类可读的方式显示天、小时、分钟等。

例如:

from datetime import datetime
start_time = datetime.now()
# do your work here
end_time = datetime.now()
print('Duration: {}'.format(end_time - start_time))

样本输出,例如。

Duration: 0:00:08.309267

or

Duration: 1 day, 1:51:24.269711

正如J.F.Sebastian所提到的,这种方法在当地时间可能会遇到一些棘手的情况,因此使用更安全:

import time
from datetime import timedelta
start_time = time.monotonic()
end_time = time.monotonic()
print(timedelta(seconds=end_time - start_time))

有一个timeit模块,可用于对Python代码的执行时间进行计时。

它在Python文档26.6中有详细的文档和示例。timeit-测量小代码片段的执行时间。


我也喜欢Paul McGuire的回答,并提出了一个更符合我需求的上下文管理器表单。

import datetime as dt
import timeit

class TimingManager(object):
    """Context Manager used with the statement 'with' to time some execution.

    Example:

    with TimingManager() as t:
       # Code to time
    """

    clock = timeit.default_timer

    def __enter__(self):
        """
        """
        self.start = self.clock()
        self.log('\n=> Start Timing: {}')

        return self

    def __exit__(self, exc_type, exc_val, exc_tb):
        """
        """
        self.endlog()

        return False

    def log(self, s, elapsed=None):
        """Log current time and elapsed time if present.
        :param s: Text to display, use '{}' to format the text with
            the current time.
        :param elapsed: Elapsed time to display. Dafault: None, no display.
        """
        print s.format(self._secondsToStr(self.clock()))

        if(elapsed is not None):
            print 'Elapsed time: {}\n'.format(elapsed)

    def endlog(self):
        """Log time for the end of execution with elapsed time.
        """
        self.log('=> End Timing: {}', self.now())

    def now(self):
        """Return current elapsed time as hh:mm:ss string.
        :return: String.
        """
        return str(dt.timedelta(seconds = self.clock() - self.start))

    def _secondsToStr(self, sec):
        """Convert timestamp to h:mm:ss string.
        :param sec: Timestamp.
        """
        return str(dt.datetime.fromtimestamp(sec))

我看过timeit模块,但它似乎只适用于小代码片段。我想给整个节目计时。

$ python -mtimeit -n1 -r1 -t -s "from your_module import main" "main()"

它运行一次your_module.main()函数,并使用time.time()函数作为计时器打印经过的时间。

要在Python中模拟/usr/bin/time,请参阅带有/usr/bin/time:如何捕获计时信息但忽略所有其他输出?的Python子进程?。

要测量每个函数的CPU时间(例如,不要包括time.sleep()期间的时间),可以使用profile模块(Python 2上的cProfile):

$ python3 -mprofile your_module.py

如果您想使用与配置文件模块相同的计时器,可以将-p传递给上面的timeit命令。

请参见如何评测Python脚本?


这是保罗·麦奎尔的回答,对我来说很有用。以防有人在运行这个问题时遇到问题。

import atexit
from time import clock

def reduce(function, iterable, initializer=None):
    it = iter(iterable)
    if initializer is None:
        value = next(it)
    else:
        value = initializer
    for element in it:
        value = function(value, element)
    return value

def secondsToStr(t):
    return "%d:%02d:%02d.%03d" % \
        reduce(lambda ll,b : divmod(ll[0],b) + ll[1:],
            [(t*1000,),1000,60,60])

line = "="*40
def log(s, elapsed=None):
    print (line)
    print (secondsToStr(clock()), '-', s)
    if elapsed:
        print ("Elapsed time:", elapsed)
    print (line)

def endlog():
    end = clock()
    elapsed = end-start
    log("End Program", secondsToStr(elapsed))

def now():
    return secondsToStr(clock())

def main():
    start = clock()
    atexit.register(endlog)
    log("Start Program")

导入文件后,从程序中调用timing.main()。


在IPython中,“timeit”任何脚本:

def foo():
    %run bar.py
timeit foo()

from time import time
start_time = time()
...
end_time = time()
time_taken = end_time - start_time # time_taken is in seconds
hours, rest = divmod(time_taken,3600)
minutes, seconds = divmod(rest, 60)

time.clock()

自3.3版起已弃用:此函数的行为取决于在平台上:改用perf_counter()或process_time(),这取决于您的需求,以具有定义良好的行为。

time.perf_counter()

返回性能计数器的值(以秒为单位),即具有最高可用分辨率的时钟来测量短路期间它包括睡眠期间的时间系统范围内。

time.process_time()

返回系统和当前进程的用户CPU时间。它不包括经过的时间在睡眠期间。

start = time.process_time()
... do something
elapsed = (time.process_time() - start)

以下代码段以可读的<HH:MM:SS>格式打印经过的时间。

import time
from datetime import timedelta

start_time = time.time()

#
# Perform lots of computations.
#

elapsed_time_secs = time.time() - start_time

msg = "Execution took: %s secs (Wall clock time)" % timedelta(seconds=round(elapsed_time_secs))

print(msg)    

要使用metakermit对Python 2.7的更新答案,您需要单调包。

代码如下:

from datetime import timedelta
from monotonic import monotonic

start_time = monotonic()
end_time = monotonic()
print(timedelta(seconds=end_time - start_time))

Python程序执行度量的时间可能不一致,具体取决于:

可以使用不同的算法评估相同的程序运行时间因算法而异运行时间因实现而异运行时间因计算机而异基于小输入,运行时间不可预测

这是因为最有效的方法是使用“增长顺序”,并学习大“O”符号来正确地执行。

无论如何,您可以尝试使用以下简单算法来评估任何Python程序在每秒特定机器计数步骤中的性能:使其适应您想要评估的计划

import time

now = time.time()
future = now + 10
step = 4 # Why 4 steps? Because until here already four operations executed
while time.time() < future:
    step += 3 # Why 3 again? Because a while loop executes one comparison and one plus equal statement
step += 4 # Why 3 more? Because one comparison starting while when time is over plus the final assignment of step + 1 and print statement
print(str(int(step / 10)) + " steps per second")

只需使用timeit模块。它同时适用于Python 2和Python 3。

import timeit

start = timeit.default_timer()

# All the program statements
stop = timeit.default_timer()
execution_time = stop - start

print("Program Executed in "+str(execution_time)) # It returns time in seconds

它在几秒钟内返回,您可以获得执行时间。这很简单,但您应该在启动程序执行的主函数中编写这些。如果您想获得执行时间,即使在出现错误时,也可以将参数“Start”设置为它,并在那里进行如下计算:

def sample_function(start,**kwargs):
     try:
         # Your statements
     except:
         # except statements run when your statements raise an exception
         stop = timeit.default_timer()
         execution_time = stop - start
         print("Program executed in " + str(execution_time))

Timeit是Python中的一个类,用于计算小代码块的执行时间。

Default_timer是此类中的一个方法,用于测量墙上时钟计时,而不是CPU执行时间。因此,其他进程执行可能会对此产生干扰。因此,它对小代码块很有用。

代码示例如下:

from timeit import default_timer as timer

start= timer()

# Some logic

end = timer()

print("Time taken:", end-start)

使用line_profiler。

line_profiler将描述单个代码行执行所需的时间。分析器通过Cython在C语言中实现,以减少分析开销。

from line_profiler import LineProfiler
import random

def do_stuff(numbers):
    s = sum(numbers)
    l = [numbers[i]/43 for i in range(len(numbers))]
    m = ['hello'+str(numbers[i]) for i in range(len(numbers))]

numbers = [random.randint(1,100) for i in range(1000)]
lp = LineProfiler()
lp_wrapper = lp(do_stuff)
lp_wrapper(numbers)
lp.print_stats()

结果将是:

Timer unit: 1e-06 s

Total time: 0.000649 s
File: <ipython-input-2-2e060b054fea>
Function: do_stuff at line 4

Line #      Hits         Time  Per Hit   % Time  Line Contents
==============================================================
     4                                           def do_stuff(numbers):
     5         1           10     10.0      1.5      s = sum(numbers)
     6         1          186    186.0     28.7      l = [numbers[i]/43 for i in range(len(numbers))]
     7         1          453    453.0     69.8      m = ['hello'+str(numbers[i]) for i in range(len(numbers))]

对于使用Jupyter笔记本的数据人员

在单元格中,可以使用Jupyter的%%time魔术命令来测量执行时间:

%%time
[ x**2 for x in range(10000)]

输出

CPU times: user 4.54 ms, sys: 0 ns, total: 4.54 ms
Wall time: 4.12 ms

这将仅捕获特定单元的执行时间。如果您想捕获整个笔记本(即程序)的执行时间,可以在同一目录中创建一个新笔记本,并在新笔记本中执行所有单元格:

假设上面的笔记本名为example_notebook.ipynb。在同一目录中的新笔记本中:

# Convert your notebook to a .py script:
!jupyter nbconvert --to script example_notebook.ipynb

# Run the example_notebook with -t flag for time
%run -t example_notebook

输出

IPython CPU timings (estimated):
  User   :       0.00 s.
  System :       0.00 s.
Wall time:       0.00 s.

我使用了一个非常简单的函数来计时代码执行的一部分:

import time
def timing():
    start_time = time.time()
    return lambda x: print("[{:.2f}s] {}".format(time.time() - start_time, x))

要使用它,只需在代码之前调用它来度量以检索函数计时,然后在代码之后调用带有注释的函数。时间将显示在评论前面。例如:

t = timing()
train = pd.read_csv('train.csv',
                        dtype={
                            'id': str,
                            'vendor_id': str,
                            'pickup_datetime': str,
                            'dropoff_datetime': str,
                            'passenger_count': int,
                            'pickup_longitude': np.float64,
                            'pickup_latitude': np.float64,
                            'dropoff_longitude': np.float64,
                            'dropoff_latitude': np.float64,
                            'store_and_fwd_flag': str,
                            'trip_duration': int,
                        },
                        parse_dates = ['pickup_datetime', 'dropoff_datetime'],
                   )
t("Loaded {} rows data from 'train'".format(len(train)))

然后输出将如下所示:

[9.35s] Loaded 1458644 rows data from 'train'

如果您想以微秒为单位测量时间,那么可以使用以下版本,完全基于Paul McGuire和Nicojo的答案——这是Python 3代码。我还为它添加了一些颜色:

import atexit
from time import time
from datetime import timedelta, datetime


def seconds_to_str(elapsed=None):
    if elapsed is None:
        return datetime.now().strftime("%Y-%m-%d %H:%M:%S.%f")
    else:
        return str(timedelta(seconds=elapsed))


def log(txt, elapsed=None):
    colour_cyan = '\033[36m'
    colour_reset = '\033[0;0;39m'
    colour_red = '\033[31m'
    print('\n ' + colour_cyan + '  [TIMING]> [' + seconds_to_str() + '] ----> ' + txt + '\n' + colour_reset)
    if elapsed:
        print("\n " + colour_red + " [TIMING]> Elapsed time ==> " + elapsed + "\n" + colour_reset)


def end_log():
    end = time()
    elapsed = end-start
    log("End Program", seconds_to_str(elapsed))


start = time()
atexit.register(end_log)
log("Start Program")

log()=>打印定时信息的函数。

txt==>要记录的第一个参数及其用于标记计时的字符串。

atexit==>Python模块,用于注册程序退出时可以调用的函数。


您只需在Python中执行此操作。没有必要让它变得复杂。

import time

start = time.localtime()
end = time.localtime()
"""Total execution time in minutes$ """
print(end.tm_min - start.tm_min)
"""Total execution time in seconds$ """
print(end.tm_sec - start.tm_sec)

这是获取程序运行时间的最简单方法:

在程序末尾编写以下代码。

import time
print(time.clock())

我在很多地方都遇到了同样的问题,所以我创建了一个方便的套装占星术。你可以用pip安装钟表,然后以优雅的方式安装:

from horology import Timing

with Timing(name='Important calculations: '):
    prepare()
    do_your_stuff()
    finish_sth()

将输出:

Important calculations: 12.43 ms

或者更简单(如果你有一个功能):

from horology import timed

@timed
def main():
    ...

将输出:

main: 7.12 h

它负责单位和舍入。它适用于python 3.6或更高版本。


稍后回答,但我使用内置的timeit:

import timeit
code_to_test = """
a = range(100000)
b = []
for i in a:
    b.append(i*2)
"""
elapsed_time = timeit.timeit(code_to_test, number=500)
print(elapsed_time)
# 10.159821493085474

在code_to_test中包装所有代码,包括可能的任何导入。number参数指定代码应该重复的次数。演示


与@rogeriopvl的响应类似,我添加了一个轻微的修改,使用相同的库将长时间运行的作业转换为小时-分钟-秒。

import time
start_time = time.time()
main()
seconds = time.time() - start_time
print('Time Taken:', time.strftime("%H:%M:%S",time.gmtime(seconds)))

样本输出

Time Taken: 00:00:08

首先,以管理员身份打开命令提示符(CMD)并在那里键入,安装人性化的软件包-pip安装人性化

代码:

from humanfriendly import format_timespan
import time
begin_time = time.time()
# Put your code here
end_time = time.time() - begin_time
print("Total execution time: ", format_timespan(end_time))

输出:


我尝试使用以下脚本找到时间差。

import time

start_time = time.perf_counter()
[main code here]
print (time.perf_counter() - start_time, "seconds")

对于函数,我建议使用我创建的这个简单的修饰符。

def timeit(method):
    def timed(*args, **kwargs):
        ts = time.time()
        result = method(*args, **kwargs)
        te = time.time()
        if 'log_time' in kwargs:
            name = kwargs.get('log_name', method.__name__.upper())
            kwargs['log_time'][name] = int((te - ts) * 1000)
        else:
            print('%r  %2.22f ms' % (method.__name__, (te - ts) * 1000))
        return result
    return timed

@timeit
def foo():
    do_some_work()

# foo()
# 'foo'  0.000953 ms

根据这个答案,创建了一个简单但方便的工具。

import time
from datetime import timedelta

def start_time_measure(message=None):
    if message:
        print(message)
    return time.monotonic()

def end_time_measure(start_time, print_prefix=None):
    end_time = time.monotonic()
    if print_prefix:
        print(print_prefix + str(timedelta(seconds=end_time - start_time)))
    return end_time

用法:

total_start_time = start_time_measure()    
start_time = start_time_measure('Doing something...')
# Do something
end_time_measure(start_time, 'Done in: ')
start_time = start_time_measure('Doing something else...')
# Do something else
end_time_measure(start_time, 'Done in: ')
end_time_measure(total_start_time, 'Total time: ')

输出:

Doing something...
Done in: 0:00:01.218000
Doing something else...
Done in: 0:00:01.313000
Total time: 0:00:02.672000

time.clock在Python 3.3中已被弃用,并将从Python 3.8中删除:请改用time.perf_counter或time.prrocess_time

import time
start_time = time.perf_counter ()
for x in range(1, 100):
    print(x)
end_time = time.perf_counter ()
print(end_time - start_time, "seconds")

我定义了以下Python装饰器:

def profile(fct):
  def wrapper(*args, **kw):
    start_time = time.time()
    ret = fct(*args, **kw)
    print("{} {} {} return {} in {} seconds".format(args[0].__class__.__name__,
                                                    args[0].__class__.__module__,
                                                    fct.__name__,
                                                    ret,
                                                    time.time() - start_time))
    return ret
  return wrapper

并将其用于函数或类/方法:

@profile
def main()
   ...

我使用来自ttictoc的tic和toc。

pip install ttictoc

然后可以在脚本中使用:

from ttictoc import tic,toc
tic()

# foo()

print(toc())

默认情况下,Linux或Unix系统(在macOS上测试)在终端上附带时间命令,您可以使用该命令运行Python脚本,并获取执行运行脚本的真实用户系统时间信息。

然而,默认输出不是很清楚(至少对我来说是这样),默认时间命令甚至不接受任何选项作为参数来格式化输出。这是因为time有两个版本——一个内置在bash中,只提供最小版本,另一个位于/usr/bin/time上。

/usr/bin/time命令接受其他参数,如-al、-h、-p和-o。我最喜欢的是-p,它在新行中显示输出,如下所示:

real 2.18
user 17.92
sys 2.71

我在查找两种不同方法的运行时间时遇到的问题,这两种方法用于查找所有<=一个数的素数。当在程序中进行用户输入时。

错误的方法

#Sample input for a number 20 
#Sample output [2, 3, 5, 7, 11, 13, 17, 19]
#Total Running time = 0.634 seconds

import time

start_time = time.time()

#Method 1 to find all the prime numbers <= a Number

# Function to check whether a number is prime or not.
def prime_no(num):
if num<2:
    return False
else:
    for i in range(2, num//2+1):
        if num % i == 0:
            return False
    return True

#To print all the values <= n
def Prime_under_num(n):
    a = [2]
    if n <2:
        print("None")
    elif n==2:
        print(2)
    else:
"Neglecting all even numbers as even numbers won't be prime in order to reduce the time complexity."
        for i in range(3, n+1, 2):   
            if prime_no(i):
                a.append(i)
        print(a)


"When Method 1 is only used outputs of running time for different inputs"
#Total Running time = 2.73761 seconds #n = 100
#Total Running time = 3.14781 seconds #n = 1000
#Total Running time = 8.69278 seconds #n = 10000
#Total Running time = 18.73701 seconds #n = 100000

#Method 2 to find all the prime numbers <= a Number

def Prime_under_num(n):
    a = [2]
    if n <2:
        print("None")
    elif n==2:
        print(2)
    else:
        for i in range(3, n+1, 2):   
            if n%i ==0:
                pass
            else:
                a.append(i)
        print(a)

"When Method 2 is only used outputs of running time for different inputs"
# Total Running time = 2.75935 seconds #n = 100
# Total Running time = 2.86332 seconds #n = 1000
# Total Running time = 4.59884 seconds #n = 10000
# Total Running time = 8.55057 seconds #n = 100000

if __name__ == "__main__" :
    n = int(input())
    Prime_under_num(n)
    print("Total Running time = {:.5f} seconds".format(time.time() - start_time))

上述所有情况下获得的不同运行时间都是错误的。对于我们正在接受输入的问题,我们必须在接受输入后才开始计时。这里,用户键入输入所花费的时间也与运行时间一起计算。

正确的方法

我们必须从开头删除start_time=time.time()并将其添加到主块中。

if __name__ == "__main__" :
    n = int(input())
    start_time = time.time()
    Prime_under_num(n)
    print("Total Running time = {:.3f} seconds".format(time.time() - start_time))

因此,两种方法单独使用时的输出如下:-

# Method 1

# Total Running time = 0.00159 seconds #n = 100
# Total Running time = 0.00506 seconds #n = 1000
# Total Running time = 0.22987 seconds #n = 10000
# Total Running time = 18.55819 seconds #n = 100000

# Method 2

# Total Running time = 0.00011 seconds #n = 100
# Total Running time = 0.00118 seconds #n = 1000
# Total Running time = 0.00302 seconds #n = 10000
# Total Running time = 0.01450 seconds #n = 100000

现在我们可以看到,与错误方法相比,总运行时间有显著差异。即使方法2在两种方法中的性能优于方法1,但第一种方法(错误方法)是错误的。


我认为这是最好和最简单的方法:

from time import monotonic

start_time = monotonic()
# something
print(f"Run time {monotonic() - start_time} seconds")

或与装饰师一起:

from time import monotonic
    
def record_time(function):
    def wrap(*args, **kwargs):
        start_time = monotonic()
        function_return = function(*args, **kwargs)
        print(f"Run time {monotonic() - start_time} seconds")
        return function_return
    return wrap

@record_time
def your_function():
    # something